Based on a Stephen King novel written five years prior, the 1987 Arnold Schwarzenegger film The Running Man portrayed a dystopian US that had devolved into an authoritarian police state and featured a television programme where convicted criminal “runners” had to escape death to avoid being killed by professional killers.

Schwarzenegger played an innocent police helicopter pilot who was framed for the murder of dozens of innocent people through the use of carefully edited documentary footage. Subsequently, his face would be digitally superimposed on another person’s body in existing video, giving the impression that he had been killed. Thirty-six years later, as the deepfake video featuring actor Rashmika Mandanna is causing concern for the Indian public and media, it’s noteworthy that this film foresaw the rise of “fake news” and “deepfakes” far in advance.

It appears that awareness of the existence of deepfakes in society has ignited as a result of Rashmika Mandanna’s case. And to a great extent. Though it was well known that deepfakes of glamorous actors, elite sportspersons, prominent politicians, and even prosperous business moguls could be found all over social media. Often by face-swapping, which involves adding the targets’ faces to other bodies in already-taken pictures or videos. Fake words are sometimes put in someone’s mouth. Consequently, the lines separating fact from fiction become hazy, uncertain, and blurry.

Beginning in the 1990s, researchers at academic institutions developed deepfake technology. However, it took some time to become more grounded. Only in 2017 did a Reddit user named “deepfakes” upload doctored porn videos on the platform, giving rise to the term.

Deepfake is a portmanteau of the terms “deep learning” and “fake.” They can now be produced with a wide variety of easily available, user-friendly and efficient apps, including those for mobile phones. And even in the absence of any knowledge of deep learning, amateurs are producing deepfakes. Technology is advancing to the point where it’s getting harder to distinguish it from “doctored.” All things considered, deepfake could be considered the “Photoshop” of the 21st century. The majority of deepfakes are pornographic. Indeed, according to a 2019 study by Amsterdam-based cybersecurity company Deeptrace, a startling 96 per cent of deepfakes were pornographic, and 99 per cent of them mapped the faces of female celebrities onto porn stars. This doesn’t mean that the remaining 4 per cent of deepfakes may be disregarded. Take a few instances.

Mark Zuckerberg said that he had “total control of billions of people’s stolen data” in 2019. Undoubtedly, most of us have seen some such unreal videos. A few months ago, deepfake images created by generative AI “Midjourney” of Donald Trump being stormily arrested or Pope Francis dressed in a puffy coat became widespread. The issue is that millions of people watch them in this era of social media, and regrettably, the majority of them seem to take them at face value. And this is where humanity is most in peril. Today, deepfakes affect everyone, from Biden to Taylor Swift, Zelensky to Elon Musk.

Consequently, deepfakes have the power to ruin sociopolitical stability, destabilise stock markets, and upset religious sensitivities. They can also affect crucial elections. That being said, these possibilities make sense. However, there are risks that are more difficult to recognise. In this era of deepfakes, it’s always easy to argue that a recorded conversation, image, or video is fake. Therefore, the issue might not be so much with the faked reality as it’s with the fact that the real reality starts to be tenable refuted.

Not always harmful

Nonetheless, deepfake isn’t always harmful. Joaquin Oliver, a Parkland shooting victim, was resurrected in 2020 using deepfake technology to convey a video message in support of the gun safety voting campaign. Fans of the late John Lennon were ecstatic to learn that deepfake technology would help Lennon’s voice team up with Paul McCartney. Deepfake technology is currently being used in an increasing number of films to resurrect deceased actors or to portray actors at different ages.

Can deepfakes really be controlled? Naturally, to some extent, through social media and internet regulation and by framing strict legal frameworks. However, aiming for total control might be unachievable.

One can’t put the genie back in the bottle, perhaps. Hence, the public needs to exercise greater caution and wisdom in the first place and not take any audio, video, or photo from social media at face value.

For our own survival in this complex juncture of civilisation, we, the people of this planet, must also run away from the deepfake realm. Not that simple, though.

The writer is Professor of Statistics, Indian Statistical Institute, Kolkata

comment COMMENT NOW